Image Segmentation in Satellite Image using Optimal Texture Measures

Abstract — Texture in high resolution satellite images requires substantial amendment in the conventional segmentation algorithms. In this paper, a satellite image is segmented using optimal texture measures. Satellite image used in this paper is a high resolution data which will provide more details of the urban areas, but it seems evident that it will create additional problems in terms of information extraction using automatic classification. This work improves the classification accuracy of intra-urban land cover types. Four texture measures are evaluated using grey-level co-occurrence matrix (GLCM). Four texture indices with six window sizes are obtained from satellite image. Principle Component Analysis (PCA) is applied to these texture measures. The resultant image is then compared with homogeneity texture feature image, obtained using 7×7 window. The per pixel classification accuracy is improved in this work by varying the window size